There is a discussion on the JUnit list about whether coverage tools are valuable.

I ran a coverage tool on a project of mine last night and found that almost all the coverage gaps were in boilerplate code. An interface required me to return false in a whole bunch of classes.

The duplication was already bothering me. The duplication plus coverage gaps bothered me enough to extract a common base class. Coverage was back up to almost 100% and I liked the new design better.

Go figure.

Read the full blog entry